| 1 |
/*------------------------------------------------------------------------- |
| 2 |
* |
| 3 |
* nodeMemoize.c |
| 4 |
* Routines to handle caching of results from parameterized nodes |
| 5 |
* |
| 6 |
* Portions Copyright (c) 2021-2025, PostgreSQL Global Development Group |
| 7 |
* Portions Copyright (c) 1994, Regents of the University of California |
| 8 |
* |
| 9 |
* |
| 10 |
* IDENTIFICATION |
| 11 |
* src/backend/executor/nodeMemoize.c |
| 12 |
* |
| 13 |
* Memoize nodes are intended to sit above parameterized nodes in the plan |
| 14 |
* tree in order to cache results from them. The intention here is that a |
| 15 |
* repeat scan with a parameter value that has already been seen by the node |
| 16 |
* can fetch tuples from the cache rather than having to re-scan the inner |
| 17 |
* node all over again. The query planner may choose to make use of one of |
| 18 |
* these when it thinks rescans for previously seen values are likely enough |
| 19 |
* to warrant adding the additional node. |
| 20 |
* |
| 21 |
* The method of cache we use is a hash table. When the cache fills, we never |
| 22 |
* spill tuples to disk, instead, we choose to evict the least recently used |
| 23 |
* cache entry from the cache. We remember the least recently used entry by |
| 24 |
* always pushing new entries and entries we look for onto the tail of a |
| 25 |
* doubly linked list. This means that older items always bubble to the top |
| 26 |
* of this LRU list. |
| 27 |
* |
| 28 |
* Sometimes our callers won't run their scans to completion. For example a |
| 29 |
* semi-join only needs to run until it finds a matching tuple, and once it |
| 30 |
* does, the join operator skips to the next outer tuple and does not execute |
| 31 |
* the inner side again on that scan. Because of this, we must keep track of |
| 32 |
* when a cache entry is complete, and by default, we know it is when we run |
| 33 |
* out of tuples to read during the scan. However, there are cases where we |
| 34 |
* can mark the cache entry as complete without exhausting the scan of all |
| 35 |
* tuples. One case is unique joins, where the join operator knows that there |
| 36 |
* will only be at most one match for any given outer tuple. In order to |
| 37 |
* support such cases we allow the "singlerow" option to be set for the cache. |
| 38 |
* This option marks the cache entry as complete after we read the first tuple |
| 39 |
* from the subnode. |
| 40 |
* |
| 41 |
* It's possible when we're filling the cache for a given set of parameters |
| 42 |
* that we're unable to free enough memory to store any more tuples. If this |
| 43 |
* happens then we'll have already evicted all other cache entries. When |
| 44 |
* caching another tuple would cause us to exceed our memory budget, we must |
| 45 |
* free the entry that we're currently populating and move the state machine |
| 46 |
* into MEMO_CACHE_BYPASS_MODE. This means that we'll not attempt to cache |
| 47 |
* any further tuples for this particular scan. We don't have the memory for |
| 48 |
* it. The state machine will be reset again on the next rescan. If the |
| 49 |
* memory requirements to cache the next parameter's tuples are less |
| 50 |
* demanding, then that may allow us to start putting useful entries back into |
| 51 |
* the cache again. |
| 52 |
* |
| 53 |
* |
| 54 |
* INTERFACE ROUTINES |
| 55 |
* ExecMemoize - lookup cache, exec subplan when not found |
| 56 |
* ExecInitMemoize - initialize node and subnodes |
| 57 |
* ExecEndMemoize - shutdown node and subnodes |
| 58 |
* ExecReScanMemoize - rescan the memoize node |
| 59 |
* |
| 60 |
* ExecMemoizeEstimate estimates DSM space needed for parallel plan |
| 61 |
* ExecMemoizeInitializeDSM initialize DSM for parallel plan |
| 62 |
* ExecMemoizeInitializeWorker attach to DSM info in parallel worker |
| 63 |
* ExecMemoizeRetrieveInstrumentation get instrumentation from worker |
| 64 |
*------------------------------------------------------------------------- |
| 65 |
*/ |
| 66 |
|
| 67 |
#include "postgres.h" |
| 68 |
|
| 69 |
#include "common/hashfn.h" |
| 70 |
#include "executor/executor.h" |
| 71 |
#include "executor/nodeMemoize.h" |
| 72 |
#include "lib/ilist.h" |
| 73 |
#include "miscadmin.h" |
| 74 |
#include "utils/datum.h" |
| 75 |
#include "utils/lsyscache.h" |
| 76 |
|
| 77 |
/* States of the ExecMemoize state machine */ |
| 78 |
#define MEMO_CACHE_LOOKUP 1 /* Attempt to perform a cache lookup */ |
| 79 |
#define MEMO_CACHE_FETCH_NEXT_TUPLE 2 /* Get another tuple from the cache */ |
| 80 |
#define MEMO_FILLING_CACHE 3 /* Read outer node to fill cache */ |
| 81 |
#define MEMO_CACHE_BYPASS_MODE 4 /* Bypass mode. Just read from our |
| 82 |
* subplan without caching anything */ |
| 83 |
#define MEMO_END_OF_SCAN 5 /* Ready for rescan */ |
| 84 |
|
| 85 |
|
| 86 |
/* Helper macros for memory accounting */ |
| 87 |
#define EMPTY_ENTRY_MEMORY_BYTES(e) (sizeof(MemoizeEntry) + \ |
| 88 |
sizeof(MemoizeKey) + \ |
| 89 |
(e)->key->params->t_len); |
| 90 |
#define CACHE_TUPLE_BYTES(t) (sizeof(MemoizeTuple) + \ |
| 91 |
(t)->mintuple->t_len) |
| 92 |
|
| 93 |
/* MemoizeTuple Stores an individually cached tuple */ |
| 94 |
typedef struct MemoizeTuple |
| 95 |
{ |
| 96 |
MinimalTuple mintuple; /* Cached tuple */ |
| 97 |
struct MemoizeTuple *next; /* The next tuple with the same parameter |
| 98 |
* values or NULL if it's the last one */ |
| 99 |
} MemoizeTuple; |
| 100 |
|
| 101 |
/* |
| 102 |
* MemoizeKey |
| 103 |
* The hash table key for cached entries plus the LRU list link |
| 104 |
*/ |
| 105 |
typedef struct MemoizeKey |
| 106 |
{ |
| 107 |
MinimalTuple params; |
| 108 |
dlist_node lru_node; /* Pointer to next/prev key in LRU list */ |
| 109 |
} MemoizeKey; |
| 110 |
|
| 111 |
/* |
| 112 |
* MemoizeEntry |
| 113 |
* The data struct that the cache hash table stores |
| 114 |
*/ |
| 115 |
typedef struct MemoizeEntry |
| 116 |
{ |
| 117 |
MemoizeKey *key; /* Hash key for hash table lookups */ |
| 118 |
MemoizeTuple *tuplehead; /* Pointer to the first tuple or NULL if no |
| 119 |
* tuples are cached for this entry */ |
| 120 |
uint32 hash; /* Hash value (cached) */ |
| 121 |
char status; /* Hash status */ |
| 122 |
bool complete; /* Did we read the outer plan to completion? */ |
| 123 |
} MemoizeEntry; |
| 124 |
|
| 125 |
|
| 126 |
#define SH_PREFIX memoize |
| 127 |
#define SH_ELEMENT_TYPE MemoizeEntry |
| 128 |
#define SH_KEY_TYPE MemoizeKey * |
| 129 |
#define SH_SCOPE static inline |
| 130 |
#define SH_DECLARE |
| 131 |
#include "lib/simplehash.h" |
| 132 |
|
| 133 |
static uint32 MemoizeHash_hash(struct memoize_hash *tb, |
| 134 |
const MemoizeKey *key); |
| 135 |
static bool MemoizeHash_equal(struct memoize_hash *tb, |
| 136 |
const MemoizeKey *key1, |
| 137 |
const MemoizeKey *key2); |
| 138 |
|
| 139 |
#define SH_PREFIX memoize |
| 140 |
#define SH_ELEMENT_TYPE MemoizeEntry |
| 141 |
#define SH_KEY_TYPE MemoizeKey * |
| 142 |
#define SH_KEY key |
| 143 |
#define SH_HASH_KEY(tb, key) MemoizeHash_hash(tb, key) |
| 144 |
#define SH_EQUAL(tb, a, b) MemoizeHash_equal(tb, a, b) |
| 145 |
#define SH_SCOPE static inline |
| 146 |
#define SH_STORE_HASH |
| 147 |
#define SH_GET_HASH(tb, a) a->hash |
| 148 |
#define SH_DEFINE |
| 149 |
#include "lib/simplehash.h" |
| 150 |
|
| 151 |
/* |
| 152 |
* MemoizeHash_hash |
| 153 |
* Hash function for simplehash hashtable. 'key' is unused here as we |
| 154 |
* require that all table lookups first populate the MemoizeState's |
| 155 |
* probeslot with the key values to be looked up. |
| 156 |
*/ |
| 157 |
static uint32 |
| 158 |
MemoizeHash_hash(struct memoize_hash *tb, const MemoizeKey *key) |
| 159 |
{ |
| 160 |
MemoizeState *mstate = (MemoizeState *) tb->private_data; |
| 161 |
ExprContext *econtext = mstate->ss.ps.ps_ExprContext; |
| 162 |
MemoryContext oldcontext; |
| 163 |
TupleTableSlot *pslot = mstate->probeslot; |
| 164 |
uint32 hashkey = 0; |
| 165 |
int numkeys = mstate->nkeys; |
| 166 |
|
| 167 |
oldcontext = MemoryContextSwitchTo(econtext->ecxt_per_tuple_memory); |
| 168 |
|
| 169 |
if (mstate->binary_mode) |
| 170 |
{ |
| 171 |
for (int i = 0; i < numkeys; i++) |
| 172 |
{ |
| 173 |
/* combine successive hashkeys by rotating */ |
| 174 |
hashkey = pg_rotate_left32(hashkey, 1); |
| 175 |
|
| 176 |
if (!pslot->tts_isnull[i]) /* treat nulls as having hash key 0 */ |
| 177 |
{ |
| 178 |
CompactAttribute *attr; |
| 179 |
uint32 hkey; |
| 180 |
|
| 181 |
attr = TupleDescCompactAttr(pslot->tts_tupleDescriptor, i); |
| 182 |
|
| 183 |
hkey = datum_image_hash(pslot->tts_values[i], attr->attbyval, attr->attlen); |
| 184 |
|
| 185 |
hashkey ^= hkey; |
| 186 |
} |
| 187 |
} |
| 188 |
} |
| 189 |
else |
| 190 |
{ |
| 191 |
FmgrInfo *hashfunctions = mstate->hashfunctions; |
| 192 |
Oid *collations = mstate->collations; |
| 193 |
|
| 194 |
for (int i = 0; i < numkeys; i++) |
| 195 |
{ |
| 196 |
/* combine successive hashkeys by rotating */ |
| 197 |
hashkey = pg_rotate_left32(hashkey, 1); |
| 198 |
|
| 199 |
if (!pslot->tts_isnull[i]) /* treat nulls as having hash key 0 */ |
| 200 |
{ |
| 201 |
uint32 hkey; |
| 202 |
|
| 203 |
hkey = DatumGetUInt32(FunctionCall1Coll(&hashfunctions[i], |
| 204 |
collations[i], pslot->tts_values[i])); |
| 205 |
hashkey ^= hkey; |
| 206 |
} |
| 207 |
} |
| 208 |
} |
| 209 |
|
| 210 |
MemoryContextSwitchTo(oldcontext); |
| 211 |
return murmurhash32(hashkey); |
| 212 |
} |
| 213 |
|
| 214 |
/* |
| 215 |
* MemoizeHash_equal |
| 216 |
* Equality function for confirming hash value matches during a hash |
| 217 |
* table lookup. 'key2' is never used. Instead the MemoizeState's |
| 218 |
* probeslot is always populated with details of what's being looked up. |
| 219 |
*/ |
| 220 |
static bool |
| 221 |
MemoizeHash_equal(struct memoize_hash *tb, const MemoizeKey *key1, |
| 222 |
const MemoizeKey *key2) |
| 223 |
{ |
| 224 |
MemoizeState *mstate = (MemoizeState *) tb->private_data; |
| 225 |
ExprContext *econtext = mstate->ss.ps.ps_ExprContext; |
| 226 |
TupleTableSlot *tslot = mstate->tableslot; |
| 227 |
TupleTableSlot *pslot = mstate->probeslot; |
| 228 |
|
| 229 |
/* probeslot should have already been prepared by prepare_probe_slot() */ |
| 230 |
ExecStoreMinimalTuple(key1->params, tslot, false); |
| 231 |
|
| 232 |
if (mstate->binary_mode) |
| 233 |
{ |
| 234 |
MemoryContext oldcontext; |
| 235 |
int numkeys = mstate->nkeys; |
| 236 |
bool match = true; |
| 237 |
|
| 238 |
oldcontext = MemoryContextSwitchTo(econtext->ecxt_per_tuple_memory); |
| 239 |
|
| 240 |
slot_getallattrs(tslot); |
| 241 |
slot_getallattrs(pslot); |
| 242 |
|
| 243 |
for (int i = 0; i < numkeys; i++) |
| 244 |
{ |
| 245 |
CompactAttribute *attr; |
| 246 |
|
| 247 |
if (tslot->tts_isnull[i] != pslot->tts_isnull[i]) |
| 248 |
{ |
| 249 |
match = false; |
| 250 |
break; |
| 251 |
} |
| 252 |
|
| 253 |
/* both NULL? they're equal */ |
| 254 |
if (tslot->tts_isnull[i]) |
| 255 |
continue; |
| 256 |
|
| 257 |
/* perform binary comparison on the two datums */ |
| 258 |
attr = TupleDescCompactAttr(tslot->tts_tupleDescriptor, i); |
| 259 |
if (!datum_image_eq(tslot->tts_values[i], pslot->tts_values[i], |
| 260 |
attr->attbyval, attr->attlen)) |
| 261 |
{ |
| 262 |
match = false; |
| 263 |
break; |
| 264 |
} |
| 265 |
} |
| 266 |
|
| 267 |
MemoryContextSwitchTo(oldcontext); |
| 268 |
return match; |
| 269 |
} |
| 270 |
else |
| 271 |
{ |
| 272 |
econtext->ecxt_innertuple = tslot; |
| 273 |
econtext->ecxt_outertuple = pslot; |
| 274 |
return ExecQual(mstate->cache_eq_expr, econtext); |
| 275 |
} |
| 276 |
} |
| 277 |
|
| 278 |
/* |
| 279 |
* Initialize the hash table to empty. The MemoizeState's hashtable field |
| 280 |
* must point to NULL. |
| 281 |
*/ |
| 282 |
static void |
| 283 |
build_hash_table(MemoizeState *mstate, uint32 size) |
| 284 |
{ |
| 285 |
Assert(mstate->hashtable == NULL); |
| 286 |
|
| 287 |
/* Make a guess at a good size when we're not given a valid size. */ |
| 288 |
if (size == 0) |
| 289 |
size = 1024; |
| 290 |
|
| 291 |
/* memoize_create will convert the size to a power of 2 */ |
| 292 |
mstate->hashtable = memoize_create(mstate->tableContext, size, mstate); |
| 293 |
} |
| 294 |
|
| 295 |
/* |
| 296 |
* prepare_probe_slot |
| 297 |
* Populate mstate's probeslot with the values from the tuple stored |
| 298 |
* in 'key'. If 'key' is NULL, then perform the population by evaluating |
| 299 |
* mstate's param_exprs. |
| 300 |
*/ |
| 301 |
static inline void |
| 302 |
prepare_probe_slot(MemoizeState *mstate, MemoizeKey *key) |
| 303 |
{ |
| 304 |
TupleTableSlot *pslot = mstate->probeslot; |
| 305 |
TupleTableSlot *tslot = mstate->tableslot; |
| 306 |
int numKeys = mstate->nkeys; |
| 307 |
|
| 308 |
ExecClearTuple(pslot); |
| 309 |
|
| 310 |
if (key == NULL) |
| 311 |
{ |
| 312 |
ExprContext *econtext = mstate->ss.ps.ps_ExprContext; |
| 313 |
MemoryContext oldcontext; |
| 314 |
|
| 315 |
oldcontext = MemoryContextSwitchTo(econtext->ecxt_per_tuple_memory); |
| 316 |
|
| 317 |
/* Set the probeslot's values based on the current parameter values */ |
| 318 |
for (int i = 0; i < numKeys; i++) |
| 319 |
pslot->tts_values[i] = ExecEvalExpr(mstate->param_exprs[i], |
| 320 |
econtext, |
| 321 |
&pslot->tts_isnull[i]); |
| 322 |
|
| 323 |
MemoryContextSwitchTo(oldcontext); |
| 324 |
} |
| 325 |
else |
| 326 |
{ |
| 327 |
/* Process the key's MinimalTuple and store the values in probeslot */ |
| 328 |
ExecStoreMinimalTuple(key->params, tslot, false); |
| 329 |
slot_getallattrs(tslot); |
| 330 |
memcpy(pslot->tts_values, tslot->tts_values, sizeof(Datum) * numKeys); |
| 331 |
memcpy(pslot->tts_isnull, tslot->tts_isnull, sizeof(bool) * numKeys); |
| 332 |
} |
| 333 |
|
| 334 |
ExecStoreVirtualTuple(pslot); |
| 335 |
} |
| 336 |
|
| 337 |
/* |
| 338 |
* entry_purge_tuples |
| 339 |
* Remove all tuples from the cache entry pointed to by 'entry'. This |
| 340 |
* leaves an empty cache entry. Also, update the memory accounting to |
| 341 |
* reflect the removal of the tuples. |
| 342 |
*/ |
| 343 |
static inline void |
| 344 |
entry_purge_tuples(MemoizeState *mstate, MemoizeEntry *entry) |
| 345 |
{ |
| 346 |
MemoizeTuple *tuple = entry->tuplehead; |
| 347 |
uint64 freed_mem = 0; |
| 348 |
|
| 349 |
while (tuple != NULL) |
| 350 |
{ |
| 351 |
MemoizeTuple *next = tuple->next; |
| 352 |
|
| 353 |
freed_mem += CACHE_TUPLE_BYTES(tuple); |
| 354 |
|
| 355 |
/* Free memory used for this tuple */ |
| 356 |
pfree(tuple->mintuple); |
| 357 |
pfree(tuple); |
| 358 |
|
| 359 |
tuple = next; |
| 360 |
} |
| 361 |
|
| 362 |
entry->complete = false; |
| 363 |
entry->tuplehead = NULL; |
| 364 |
|
| 365 |
/* Update the memory accounting */ |
| 366 |
mstate->mem_used -= freed_mem; |
| 367 |
} |
| 368 |
|
| 369 |
/* |
| 370 |
* remove_cache_entry |
| 371 |
* Remove 'entry' from the cache and free memory used by it. |
| 372 |
*/ |
| 373 |
static void |
| 374 |
remove_cache_entry(MemoizeState *mstate, MemoizeEntry *entry) |
| 375 |
{ |
| 376 |
MemoizeKey *key = entry->key; |
| 377 |
|
| 378 |
dlist_delete(&entry->key->lru_node); |
| 379 |
|
| 380 |
/* Remove all of the tuples from this entry */ |
| 381 |
entry_purge_tuples(mstate, entry); |
| 382 |
|
| 383 |
/* |
| 384 |
* Update memory accounting. entry_purge_tuples should have already |
| 385 |
* subtracted the memory used for each cached tuple. Here we just update |
| 386 |
* the amount used by the entry itself. |
| 387 |
*/ |
| 388 |
mstate->mem_used -= EMPTY_ENTRY_MEMORY_BYTES(entry); |
| 389 |
|
| 390 |
/* Remove the entry from the cache */ |
| 391 |
memoize_delete_item(mstate->hashtable, entry); |
| 392 |
|
| 393 |
pfree(key->params); |
| 394 |
pfree(key); |
| 395 |
} |
| 396 |
|
| 397 |
/* |
| 398 |
* cache_purge_all |
| 399 |
* Remove all items from the cache |
| 400 |
*/ |
| 401 |
static void |
| 402 |
cache_purge_all(MemoizeState *mstate) |
| 403 |
{ |
| 404 |
uint64 evictions = 0; |
| 405 |
|
| 406 |
if (mstate->hashtable != NULL) |
| 407 |
evictions = mstate->hashtable->members; |
| 408 |
|
| 409 |
/* |
| 410 |
* Likely the most efficient way to remove all items is to just reset the |
| 411 |
* memory context for the cache and then rebuild a fresh hash table. This |
| 412 |
* saves having to remove each item one by one and pfree each cached tuple |
| 413 |
*/ |
| 414 |
MemoryContextReset(mstate->tableContext); |
| 415 |
|
| 416 |
/* NULLify so we recreate the table on the next call */ |
| 417 |
mstate->hashtable = NULL; |
| 418 |
|
| 419 |
/* reset the LRU list */ |
| 420 |
dlist_init(&mstate->lru_list); |
| 421 |
mstate->last_tuple = NULL; |
| 422 |
mstate->entry = NULL; |
| 423 |
|
| 424 |
mstate->mem_used = 0; |
| 425 |
|
| 426 |
/* XXX should we add something new to track these purges? */ |
| 427 |
mstate->stats.cache_evictions += evictions; /* Update Stats */ |
| 428 |
} |
| 429 |
|
| 430 |
/* |
| 431 |
* cache_reduce_memory |
| 432 |
* Evict older and less recently used items from the cache in order to |
| 433 |
* reduce the memory consumption back to something below the |
| 434 |
* MemoizeState's mem_limit. |
| 435 |
* |
| 436 |
* 'specialkey', if not NULL, causes the function to return false if the entry |
| 437 |
* which the key belongs to is removed from the cache. |
| 438 |
*/ |
| 439 |
static bool |
| 440 |
cache_reduce_memory(MemoizeState *mstate, MemoizeKey *specialkey) |
| 441 |
{ |
| 442 |
bool specialkey_intact = true; /* for now */ |
| 443 |
dlist_mutable_iter iter; |
| 444 |
uint64 evictions = 0; |
| 445 |
|
| 446 |
/* Update peak memory usage */ |
| 447 |
if (mstate->mem_used > mstate->stats.mem_peak) |
| 448 |
mstate->stats.mem_peak = mstate->mem_used; |
| 449 |
|
| 450 |
/* We expect only to be called when we've gone over budget on memory */ |
| 451 |
Assert(mstate->mem_used > mstate->mem_limit); |
| 452 |
|
| 453 |
/* Start the eviction process starting at the head of the LRU list. */ |
| 454 |
dlist_foreach_modify(iter, &mstate->lru_list) |
| 455 |
{ |
| 456 |
MemoizeKey *key = dlist_container(MemoizeKey, lru_node, iter.cur); |
| 457 |
MemoizeEntry *entry; |
| 458 |
|
| 459 |
/* |
| 460 |
* Populate the hash probe slot in preparation for looking up this LRU |
| 461 |
* entry. |
| 462 |
*/ |
| 463 |
prepare_probe_slot(mstate, key); |
| 464 |
|
| 465 |
/* |
| 466 |
* Ideally the LRU list pointers would be stored in the entry itself |
| 467 |
* rather than in the key. Unfortunately, we can't do that as the |
| 468 |
* simplehash.h code may resize the table and allocate new memory for |
| 469 |
* entries which would result in those pointers pointing to the old |
| 470 |
* buckets. However, it's fine to use the key to store this as that's |
| 471 |
* only referenced by a pointer in the entry, which of course follows |
| 472 |
* the entry whenever the hash table is resized. Since we only have a |
| 473 |
* pointer to the key here, we must perform a hash table lookup to |
| 474 |
* find the entry that the key belongs to. |
| 475 |
*/ |
| 476 |
entry = memoize_lookup(mstate->hashtable, NULL); |
| 477 |
|
| 478 |
/* |
| 479 |
* Sanity check that we found the entry belonging to the LRU list |
| 480 |
* item. A misbehaving hash or equality function could cause the |
| 481 |
* entry not to be found or the wrong entry to be found. |
| 482 |
*/ |
| 483 |
if (unlikely(entry == NULL || entry->key != key)) |
| 484 |
elog(ERROR, "could not find memoization table entry"); |
| 485 |
|
| 486 |
/* |
| 487 |
* If we're being called to free memory while the cache is being |
| 488 |
* populated with new tuples, then we'd better take some care as we |
| 489 |
* could end up freeing the entry which 'specialkey' belongs to. |
| 490 |
* Generally callers will pass 'specialkey' as the key for the cache |
| 491 |
* entry which is currently being populated, so we must set |
| 492 |
* 'specialkey_intact' to false to inform the caller the specialkey |
| 493 |
* entry has been removed. |
| 494 |
*/ |
| 495 |
if (key == specialkey) |
| 496 |
specialkey_intact = false; |
| 497 |
|
| 498 |
/* |
| 499 |
* Finally remove the entry. This will remove from the LRU list too. |
| 500 |
*/ |
| 501 |
remove_cache_entry(mstate, entry); |
| 502 |
|
| 503 |
evictions++; |
| 504 |
|
| 505 |
/* Exit if we've freed enough memory */ |
| 506 |
if (mstate->mem_used <= mstate->mem_limit) |
| 507 |
break; |
| 508 |
} |
| 509 |
|
| 510 |
mstate->stats.cache_evictions += evictions; /* Update Stats */ |
| 511 |
|
| 512 |
return specialkey_intact; |
| 513 |
} |
| 514 |
|
| 515 |
/* |
| 516 |
* cache_lookup |
| 517 |
* Perform a lookup to see if we've already cached tuples based on the |
| 518 |
* scan's current parameters. If we find an existing entry we move it to |
| 519 |
* the end of the LRU list, set *found to true then return it. If we |
| 520 |
* don't find an entry then we create a new one and add it to the end of |
| 521 |
* the LRU list. We also update cache memory accounting and remove older |
| 522 |
* entries if we go over the memory budget. If we managed to free enough |
| 523 |
* memory we return the new entry, else we return NULL. |
| 524 |
* |
| 525 |
* Callers can assume we'll never return NULL when *found is true. |
| 526 |
*/ |
| 527 |
static MemoizeEntry * |
| 528 |
cache_lookup(MemoizeState *mstate, bool *found) |
| 529 |
{ |
| 530 |
MemoizeKey *key; |
| 531 |
MemoizeEntry *entry; |
| 532 |
MemoryContext oldcontext; |
| 533 |
|
| 534 |
/* prepare the probe slot with the current scan parameters */ |
| 535 |
prepare_probe_slot(mstate, NULL); |
| 536 |
|
| 537 |
/* |
| 538 |
* Add the new entry to the cache. No need to pass a valid key since the |
| 539 |
* hash function uses mstate's probeslot, which we populated above. |
| 540 |
*/ |
| 541 |
entry = memoize_insert(mstate->hashtable, NULL, found); |
| 542 |
|
| 543 |
if (*found) |
| 544 |
{ |
| 545 |
/* |
| 546 |
* Move existing entry to the tail of the LRU list to mark it as the |
| 547 |
* most recently used item. |
| 548 |
*/ |
| 549 |
dlist_move_tail(&mstate->lru_list, &entry->key->lru_node); |
| 550 |
|
| 551 |
return entry; |
| 552 |
} |
| 553 |
|
| 554 |
oldcontext = MemoryContextSwitchTo(mstate->tableContext); |
| 555 |
|
| 556 |
/* Allocate a new key */ |
| 557 |
entry->key = key = (MemoizeKey *) palloc(sizeof(MemoizeKey)); |
| 558 |
key->params = ExecCopySlotMinimalTuple(mstate->probeslot); |
| 559 |
|
| 560 |
/* Update the total cache memory utilization */ |
| 561 |
mstate->mem_used += EMPTY_ENTRY_MEMORY_BYTES(entry); |
| 562 |
|
| 563 |
/* Initialize this entry */ |
| 564 |
entry->complete = false; |
| 565 |
entry->tuplehead = NULL; |
| 566 |
|
| 567 |
/* |
| 568 |
* Since this is the most recently used entry, push this entry onto the |
| 569 |
* end of the LRU list. |
| 570 |
*/ |
| 571 |
dlist_push_tail(&mstate->lru_list, &entry->key->lru_node); |
| 572 |
|
| 573 |
mstate->last_tuple = NULL; |
| 574 |
|
| 575 |
MemoryContextSwitchTo(oldcontext); |
| 576 |
|
| 577 |
/* |
| 578 |
* If we've gone over our memory budget, then we'll free up some space in |
| 579 |
* the cache. |
| 580 |
*/ |
| 581 |
if (mstate->mem_used > mstate->mem_limit) |
| 582 |
{ |
| 583 |
/* |
| 584 |
* Try to free up some memory. It's highly unlikely that we'll fail |
| 585 |
* to do so here since the entry we've just added is yet to contain |
| 586 |
* any tuples and we're able to remove any other entry to reduce the |
| 587 |
* memory consumption. |
| 588 |
*/ |
| 589 |
if (unlikely(!cache_reduce_memory(mstate, key))) |
| 590 |
return NULL; |
| 591 |
|
| 592 |
/* |
| 593 |
* The process of removing entries from the cache may have caused the |
| 594 |
* code in simplehash.h to shuffle elements to earlier buckets in the |
| 595 |
* hash table. If it has, we'll need to find the entry again by |
| 596 |
* performing a lookup. Fortunately, we can detect if this has |
| 597 |
* happened by seeing if the entry is still in use and that the key |
| 598 |
* pointer matches our expected key. |
| 599 |
*/ |
| 600 |
if (entry->status != memoize_SH_IN_USE || entry->key != key) |
| 601 |
{ |
| 602 |
/* |
| 603 |
* We need to repopulate the probeslot as lookups performed during |
| 604 |
* the cache evictions above will have stored some other key. |
| 605 |
*/ |
| 606 |
prepare_probe_slot(mstate, key); |
| 607 |
|
| 608 |
/* Re-find the newly added entry */ |
| 609 |
entry = memoize_lookup(mstate->hashtable, NULL); |
| 610 |
Assert(entry != NULL); |
| 611 |
} |
| 612 |
} |
| 613 |
|
| 614 |
return entry; |
| 615 |
} |
| 616 |
|
| 617 |
/* |
| 618 |
* cache_store_tuple |
| 619 |
* Add the tuple stored in 'slot' to the mstate's current cache entry. |
| 620 |
* The cache entry must have already been made with cache_lookup(). |
| 621 |
* mstate's last_tuple field must point to the tail of mstate->entry's |
| 622 |
* list of tuples. |
| 623 |
*/ |
| 624 |
static bool |
| 625 |
cache_store_tuple(MemoizeState *mstate, TupleTableSlot *slot) |
| 626 |
{ |
| 627 |
MemoizeTuple *tuple; |
| 628 |
MemoizeEntry *entry = mstate->entry; |
| 629 |
MemoryContext oldcontext; |
| 630 |
|
| 631 |
Assert(slot != NULL); |
| 632 |
Assert(entry != NULL); |
| 633 |
|
| 634 |
oldcontext = MemoryContextSwitchTo(mstate->tableContext); |
| 635 |
|
| 636 |
tuple = (MemoizeTuple *) palloc(sizeof(MemoizeTuple)); |
| 637 |
tuple->mintuple = ExecCopySlotMinimalTuple(slot); |
| 638 |
tuple->next = NULL; |
| 639 |
|
| 640 |
/* Account for the memory we just consumed */ |
| 641 |
mstate->mem_used += CACHE_TUPLE_BYTES(tuple); |
| 642 |
|
| 643 |
if (entry->tuplehead == NULL) |
| 644 |
{ |
| 645 |
/* |
| 646 |
* This is the first tuple for this entry, so just point the list head |
| 647 |
* to it. |
| 648 |
*/ |
| 649 |
entry->tuplehead = tuple; |
| 650 |
} |
| 651 |
else |
| 652 |
{ |
| 653 |
/* push this tuple onto the tail of the list */ |
| 654 |
mstate->last_tuple->next = tuple; |
| 655 |
} |
| 656 |
|
| 657 |
mstate->last_tuple = tuple; |
| 658 |
MemoryContextSwitchTo(oldcontext); |
| 659 |
|
| 660 |
/* |
| 661 |
* If we've gone over our memory budget then free up some space in the |
| 662 |
* cache. |
| 663 |
*/ |
| 664 |
if (mstate->mem_used > mstate->mem_limit) |
| 665 |
{ |
| 666 |
MemoizeKey *key = entry->key; |
| 667 |
|
| 668 |
if (!cache_reduce_memory(mstate, key)) |
| 669 |
return false; |
| 670 |
|
| 671 |
/* |
| 672 |
* The process of removing entries from the cache may have caused the |
| 673 |
* code in simplehash.h to shuffle elements to earlier buckets in the |
| 674 |
* hash table. If it has, we'll need to find the entry again by |
| 675 |
* performing a lookup. Fortunately, we can detect if this has |
| 676 |
* happened by seeing if the entry is still in use and that the key |
| 677 |
* pointer matches our expected key. |
| 678 |
*/ |
| 679 |
if (entry->status != memoize_SH_IN_USE || entry->key != key) |
| 680 |
{ |
| 681 |
/* |
| 682 |
* We need to repopulate the probeslot as lookups performed during |
| 683 |
* the cache evictions above will have stored some other key. |
| 684 |
*/ |
| 685 |
prepare_probe_slot(mstate, key); |
| 686 |
|
| 687 |
/* Re-find the entry */ |
| 688 |
mstate->entry = entry = memoize_lookup(mstate->hashtable, NULL); |
| 689 |
Assert(entry != NULL); |
| 690 |
} |
| 691 |
} |
| 692 |
|
| 693 |
return true; |
| 694 |
} |
| 695 |
|
| 696 |
static TupleTableSlot * |
| 697 |
ExecMemoize(PlanState *pstate) |
| 698 |
{ |
| 699 |
MemoizeState *node = castNode(MemoizeState, pstate); |
| 700 |
ExprContext *econtext = node->ss.ps.ps_ExprContext; |
| 701 |
PlanState *outerNode; |
| 702 |
TupleTableSlot *slot; |
| 703 |
|
| 704 |
CHECK_FOR_INTERRUPTS(); |
| 705 |
|
| 706 |
/* |
| 707 |
* Reset per-tuple memory context to free any expression evaluation |
| 708 |
* storage allocated in the previous tuple cycle. |
| 709 |
*/ |
| 710 |
ResetExprContext(econtext); |
| 711 |
|
| 712 |
switch (node->mstatus) |
| 713 |
{ |
| 714 |
case MEMO_CACHE_LOOKUP: |
| 715 |
{ |
| 716 |
MemoizeEntry *entry; |
| 717 |
TupleTableSlot *outerslot; |
| 718 |
bool found; |
| 719 |
|
| 720 |
Assert(node->entry == NULL); |
| 721 |
|
| 722 |
/* first call? we'll need a hash table. */ |
| 723 |
if (unlikely(node->hashtable == NULL)) |
| 724 |
build_hash_table(node, ((Memoize *) pstate->plan)->est_entries); |
| 725 |
|
| 726 |
/* |
| 727 |
* We're only ever in this state for the first call of the |
| 728 |
* scan. Here we have a look to see if we've already seen the |
| 729 |
* current parameters before and if we have already cached a |
| 730 |
* complete set of records that the outer plan will return for |
| 731 |
* these parameters. |
| 732 |
* |
| 733 |
* When we find a valid cache entry, we'll return the first |
| 734 |
* tuple from it. If not found, we'll create a cache entry and |
| 735 |
* then try to fetch a tuple from the outer scan. If we find |
| 736 |
* one there, we'll try to cache it. |
| 737 |
*/ |
| 738 |
|
| 739 |
/* see if we've got anything cached for the current parameters */ |
| 740 |
entry = cache_lookup(node, &found); |
| 741 |
|
| 742 |
if (found && entry->complete) |
| 743 |
{ |
| 744 |
node->stats.cache_hits += 1; /* stats update */ |
| 745 |
|
| 746 |
/* |
| 747 |
* Set last_tuple and entry so that the state |
| 748 |
* MEMO_CACHE_FETCH_NEXT_TUPLE can easily find the next |
| 749 |
* tuple for these parameters. |
| 750 |
*/ |
| 751 |
node->last_tuple = entry->tuplehead; |
| 752 |
node->entry = entry; |
| 753 |
|
| 754 |
/* Fetch the first cached tuple, if there is one */ |
| 755 |
if (entry->tuplehead) |
| 756 |
{ |
| 757 |
node->mstatus = MEMO_CACHE_FETCH_NEXT_TUPLE; |
| 758 |
|
| 759 |
slot = node->ss.ps.ps_ResultTupleSlot; |
| 760 |
ExecStoreMinimalTuple(entry->tuplehead->mintuple, |
| 761 |
slot, false); |
| 762 |
|
| 763 |
return slot; |
| 764 |
} |
| 765 |
|
| 766 |
/* The cache entry is void of any tuples. */ |
| 767 |
node->mstatus = MEMO_END_OF_SCAN; |
| 768 |
return NULL; |
| 769 |
} |
| 770 |
|
| 771 |
/* Handle cache miss */ |
| 772 |
node->stats.cache_misses += 1; /* stats update */ |
| 773 |
|
| 774 |
if (found) |
| 775 |
{ |
| 776 |
/* |
| 777 |
* A cache entry was found, but the scan for that entry |
| 778 |
* did not run to completion. We'll just remove all |
| 779 |
* tuples and start again. It might be tempting to |
| 780 |
* continue where we left off, but there's no guarantee |
| 781 |
* the outer node will produce the tuples in the same |
| 782 |
* order as it did last time. |
| 783 |
*/ |
| 784 |
entry_purge_tuples(node, entry); |
| 785 |
} |
| 786 |
|
| 787 |
/* Scan the outer node for a tuple to cache */ |
| 788 |
outerNode = outerPlanState(node); |
| 789 |
outerslot = ExecProcNode(outerNode); |
| 790 |
if (TupIsNull(outerslot)) |
| 791 |
{ |
| 792 |
/* |
| 793 |
* cache_lookup may have returned NULL due to failure to |
| 794 |
* free enough cache space, so ensure we don't do anything |
| 795 |
* here that assumes it worked. There's no need to go into |
| 796 |
* bypass mode here as we're setting mstatus to end of |
| 797 |
* scan. |
| 798 |
*/ |
| 799 |
if (likely(entry)) |
| 800 |
entry->complete = true; |
| 801 |
|
| 802 |
node->mstatus = MEMO_END_OF_SCAN; |
| 803 |
return NULL; |
| 804 |
} |
| 805 |
|
| 806 |
node->entry = entry; |
| 807 |
|
| 808 |
/* |
| 809 |
* If we failed to create the entry or failed to store the |
| 810 |
* tuple in the entry, then go into bypass mode. |
| 811 |
*/ |
| 812 |
if (unlikely(entry == NULL || |
| 813 |
!cache_store_tuple(node, outerslot))) |
| 814 |
{ |
| 815 |
node->stats.cache_overflows += 1; /* stats update */ |
| 816 |
|
| 817 |
node->mstatus = MEMO_CACHE_BYPASS_MODE; |
| 818 |
|
| 819 |
/* |
| 820 |
* No need to clear out last_tuple as we'll stay in bypass |
| 821 |
* mode until the end of the scan. |
| 822 |
*/ |
| 823 |
} |
| 824 |
else |
| 825 |
{ |
| 826 |
/* |
| 827 |
* If we only expect a single row from this scan then we |
| 828 |
* can mark that we're not expecting more. This allows |
| 829 |
* cache lookups to work even when the scan has not been |
| 830 |
* executed to completion. |
| 831 |
*/ |
| 832 |
entry->complete = node->singlerow; |
| 833 |
node->mstatus = MEMO_FILLING_CACHE; |
| 834 |
} |
| 835 |
|
| 836 |
slot = node->ss.ps.ps_ResultTupleSlot; |
| 837 |
ExecCopySlot(slot, outerslot); |
| 838 |
return slot; |
| 839 |
} |
| 840 |
|
| 841 |
case MEMO_CACHE_FETCH_NEXT_TUPLE: |
| 842 |
{ |
| 843 |
/* We shouldn't be in this state if these are not set */ |
| 844 |
Assert(node->entry != NULL); |
| 845 |
Assert(node->last_tuple != NULL); |
| 846 |
|
| 847 |
/* Skip to the next tuple to output */ |
| 848 |
node->last_tuple = node->last_tuple->next; |
| 849 |
|
| 850 |
/* No more tuples in the cache */ |
| 851 |
if (node->last_tuple == NULL) |
| 852 |
{ |
| 853 |
node->mstatus = MEMO_END_OF_SCAN; |
| 854 |
return NULL; |
| 855 |
} |
| 856 |
|
| 857 |
slot = node->ss.ps.ps_ResultTupleSlot; |
| 858 |
ExecStoreMinimalTuple(node->last_tuple->mintuple, slot, |
| 859 |
false); |
| 860 |
|
| 861 |
return slot; |
| 862 |
} |
| 863 |
|
| 864 |
case MEMO_FILLING_CACHE: |
| 865 |
{ |
| 866 |
TupleTableSlot *outerslot; |
| 867 |
MemoizeEntry *entry = node->entry; |
| 868 |
|
| 869 |
/* entry should already have been set by MEMO_CACHE_LOOKUP */ |
| 870 |
Assert(entry != NULL); |
| 871 |
|
| 872 |
/* |
| 873 |
* When in the MEMO_FILLING_CACHE state, we've just had a |
| 874 |
* cache miss and are populating the cache with the current |
| 875 |
* scan tuples. |
| 876 |
*/ |
| 877 |
outerNode = outerPlanState(node); |
| 878 |
outerslot = ExecProcNode(outerNode); |
| 879 |
if (TupIsNull(outerslot)) |
| 880 |
{ |
| 881 |
/* No more tuples. Mark it as complete */ |
| 882 |
entry->complete = true; |
| 883 |
node->mstatus = MEMO_END_OF_SCAN; |
| 884 |
return NULL; |
| 885 |
} |
| 886 |
|
| 887 |
/* |
| 888 |
* Validate if the planner properly set the singlerow flag. It |
| 889 |
* should only set that if each cache entry can, at most, |
| 890 |
* return 1 row. |
| 891 |
*/ |
| 892 |
if (unlikely(entry->complete)) |
| 893 |
elog(ERROR, "cache entry already complete"); |
| 894 |
|
| 895 |
/* Record the tuple in the current cache entry */ |
| 896 |
if (unlikely(!cache_store_tuple(node, outerslot))) |
| 897 |
{ |
| 898 |
/* Couldn't store it? Handle overflow */ |
| 899 |
node->stats.cache_overflows += 1; /* stats update */ |
| 900 |
|
| 901 |
node->mstatus = MEMO_CACHE_BYPASS_MODE; |
| 902 |
|
| 903 |
/* |
| 904 |
* No need to clear out entry or last_tuple as we'll stay |
| 905 |
* in bypass mode until the end of the scan. |
| 906 |
*/ |
| 907 |
} |
| 908 |
|
| 909 |
slot = node->ss.ps.ps_ResultTupleSlot; |
| 910 |
ExecCopySlot(slot, outerslot); |
| 911 |
return slot; |
| 912 |
} |
| 913 |
|
| 914 |
case MEMO_CACHE_BYPASS_MODE: |
| 915 |
{ |
| 916 |
TupleTableSlot *outerslot; |
| 917 |
|
| 918 |
/* |
| 919 |
* When in bypass mode we just continue to read tuples without |
| 920 |
* caching. We need to wait until the next rescan before we |
| 921 |
* can come out of this mode. |
| 922 |
*/ |
| 923 |
outerNode = outerPlanState(node); |
| 924 |
outerslot = ExecProcNode(outerNode); |
| 925 |
if (TupIsNull(outerslot)) |
| 926 |
{ |
| 927 |
node->mstatus = MEMO_END_OF_SCAN; |
| 928 |
return NULL; |
| 929 |
} |
| 930 |
|
| 931 |
slot = node->ss.ps.ps_ResultTupleSlot; |
| 932 |
ExecCopySlot(slot, outerslot); |
| 933 |
return slot; |
| 934 |
} |
| 935 |
|
| 936 |
case MEMO_END_OF_SCAN: |
| 937 |
|
| 938 |
/* |
| 939 |
* We've already returned NULL for this scan, but just in case |
| 940 |
* something calls us again by mistake. |
| 941 |
*/ |
| 942 |
return NULL; |
| 943 |
|
| 944 |
default: |
| 945 |
elog(ERROR, "unrecognized memoize state: %d", |
| 946 |
(int) node->mstatus); |
| 947 |
return NULL; |
| 948 |
} /* switch */ |
| 949 |
} |
| 950 |
|
| 951 |
MemoizeState * |
| 952 |
ExecInitMemoize(Memoize *node, EState *estate, int eflags) |
| 953 |
{ |
| 954 |
MemoizeState *mstate = makeNode(MemoizeState); |
| 955 |
Plan *outerNode; |
| 956 |
int i; |
| 957 |
int nkeys; |
| 958 |
Oid *eqfuncoids; |
| 959 |
|
| 960 |
/* check for unsupported flags */ |
| 961 |
Assert(!(eflags & (EXEC_FLAG_BACKWARD | EXEC_FLAG_MARK))); |
| 962 |
|
| 963 |
mstate->ss.ps.plan = (Plan *) node; |
| 964 |
mstate->ss.ps.state = estate; |
| 965 |
mstate->ss.ps.ExecProcNode = ExecMemoize; |
| 966 |
|
| 967 |
/* |
| 968 |
* Miscellaneous initialization |
| 969 |
* |
| 970 |
* create expression context for node |
| 971 |
*/ |
| 972 |
ExecAssignExprContext(estate, &mstate->ss.ps); |
| 973 |
|
| 974 |
outerNode = outerPlan(node); |
| 975 |
outerPlanState(mstate) = ExecInitNode(outerNode, estate, eflags); |
| 976 |
|
| 977 |
/* |
| 978 |
* Initialize return slot and type. No need to initialize projection info |
| 979 |
* because this node doesn't do projections. |
| 980 |
*/ |
| 981 |
ExecInitResultTupleSlotTL(&mstate->ss.ps, &TTSOpsMinimalTuple); |
| 982 |
mstate->ss.ps.ps_ProjInfo = NULL; |
| 983 |
|
| 984 |
/* |
| 985 |
* Initialize scan slot and type. |
| 986 |
*/ |
| 987 |
ExecCreateScanSlotFromOuterPlan(estate, &mstate->ss, &TTSOpsMinimalTuple); |
| 988 |
|
| 989 |
/* |
| 990 |
* Set the state machine to lookup the cache. We won't find anything |
| 991 |
* until we cache something, but this saves a special case to create the |
| 992 |
* first entry. |
| 993 |
*/ |
| 994 |
mstate->mstatus = MEMO_CACHE_LOOKUP; |
| 995 |
|
| 996 |
mstate->nkeys = nkeys = node->numKeys; |
| 997 |
mstate->hashkeydesc = ExecTypeFromExprList(node->param_exprs); |
| 998 |
mstate->tableslot = MakeSingleTupleTableSlot(mstate->hashkeydesc, |
| 999 |
&TTSOpsMinimalTuple); |
| 1000 |
mstate->probeslot = MakeSingleTupleTableSlot(mstate->hashkeydesc, |
| 1001 |
&TTSOpsVirtual); |
| 1002 |
|
| 1003 |
mstate->param_exprs = (ExprState **) palloc(nkeys * sizeof(ExprState *)); |
| 1004 |
mstate->collations = node->collations; /* Just point directly to the plan |
| 1005 |
* data */ |
| 1006 |
mstate->hashfunctions = (FmgrInfo *) palloc(nkeys * sizeof(FmgrInfo)); |
| 1007 |
|
| 1008 |
eqfuncoids = palloc(nkeys * sizeof(Oid)); |
| 1009 |
|
| 1010 |
for (i = 0; i < nkeys; i++) |
| 1011 |
{ |
| 1012 |
Oid hashop = node->hashOperators[i]; |
| 1013 |
Oid left_hashfn; |
| 1014 |
Oid right_hashfn; |
| 1015 |
Expr *param_expr = (Expr *) list_nth(node->param_exprs, i); |
| 1016 |
|
| 1017 |
if (!get_op_hash_functions(hashop, &left_hashfn, &right_hashfn)) |
| 1018 |
elog(ERROR, "could not find hash function for hash operator %u", |
| 1019 |
hashop); |
| 1020 |
|
| 1021 |
fmgr_info(left_hashfn, &mstate->hashfunctions[i]); |
| 1022 |
|
| 1023 |
mstate->param_exprs[i] = ExecInitExpr(param_expr, (PlanState *) mstate); |
| 1024 |
eqfuncoids[i] = get_opcode(hashop); |
| 1025 |
} |
| 1026 |
|
| 1027 |
mstate->cache_eq_expr = ExecBuildParamSetEqual(mstate->hashkeydesc, |
| 1028 |
&TTSOpsMinimalTuple, |
| 1029 |
&TTSOpsVirtual, |
| 1030 |
eqfuncoids, |
| 1031 |
node->collations, |
| 1032 |
node->param_exprs, |
| 1033 |
(PlanState *) mstate); |
| 1034 |
|
| 1035 |
pfree(eqfuncoids); |
| 1036 |
mstate->mem_used = 0; |
| 1037 |
|
| 1038 |
/* Limit the total memory consumed by the cache to this */ |
| 1039 |
mstate->mem_limit = get_hash_memory_limit(); |
| 1040 |
|
| 1041 |
/* A memory context dedicated for the cache */ |
| 1042 |
mstate->tableContext = AllocSetContextCreate(CurrentMemoryContext, |
| 1043 |
"MemoizeHashTable", |
| 1044 |
ALLOCSET_DEFAULT_SIZES); |
| 1045 |
|
| 1046 |
dlist_init(&mstate->lru_list); |
| 1047 |
mstate->last_tuple = NULL; |
| 1048 |
mstate->entry = NULL; |
| 1049 |
|
| 1050 |
/* |
| 1051 |
* Mark if we can assume the cache entry is completed after we get the |
| 1052 |
* first record for it. Some callers might not call us again after |
| 1053 |
* getting the first match. e.g. A join operator performing a unique join |
| 1054 |
* is able to skip to the next outer tuple after getting the first |
| 1055 |
* matching inner tuple. In this case, the cache entry is complete after |
| 1056 |
* getting the first tuple. This allows us to mark it as so. |
| 1057 |
*/ |
| 1058 |
mstate->singlerow = node->singlerow; |
| 1059 |
mstate->keyparamids = node->keyparamids; |
| 1060 |
|
| 1061 |
/* |
| 1062 |
* Record if the cache keys should be compared bit by bit, or logically |
| 1063 |
* using the type's hash equality operator |
| 1064 |
*/ |
| 1065 |
mstate->binary_mode = node->binary_mode; |
| 1066 |
|
| 1067 |
/* Zero the statistics counters */ |
| 1068 |
memset(&mstate->stats, 0, sizeof(MemoizeInstrumentation)); |
| 1069 |
|
| 1070 |
/* |
| 1071 |
* Because it may require a large allocation, we delay building of the |
| 1072 |
* hash table until executor run. |
| 1073 |
*/ |
| 1074 |
mstate->hashtable = NULL; |
| 1075 |
|
| 1076 |
return mstate; |
| 1077 |
} |
| 1078 |
|
| 1079 |
void |
| 1080 |
ExecEndMemoize(MemoizeState *node) |
| 1081 |
{ |
| 1082 |
#ifdef USE_ASSERT_CHECKING |
| 1083 |
/* Validate the memory accounting code is correct in assert builds. */ |
| 1084 |
if (node->hashtable != NULL) |
| 1085 |
{ |
| 1086 |
int count; |
| 1087 |
uint64 mem = 0; |
| 1088 |
memoize_iterator i; |
| 1089 |
MemoizeEntry *entry; |
| 1090 |
|
| 1091 |
memoize_start_iterate(node->hashtable, &i); |
| 1092 |
|
| 1093 |
count = 0; |
| 1094 |
while ((entry = memoize_iterate(node->hashtable, &i)) != NULL) |
| 1095 |
{ |
| 1096 |
MemoizeTuple *tuple = entry->tuplehead; |
| 1097 |
|
| 1098 |
mem += EMPTY_ENTRY_MEMORY_BYTES(entry); |
| 1099 |
while (tuple != NULL) |
| 1100 |
{ |
| 1101 |
mem += CACHE_TUPLE_BYTES(tuple); |
| 1102 |
tuple = tuple->next; |
| 1103 |
} |
| 1104 |
count++; |
| 1105 |
} |
| 1106 |
|
| 1107 |
Assert(count == node->hashtable->members); |
| 1108 |
Assert(mem == node->mem_used); |
| 1109 |
} |
| 1110 |
#endif |
| 1111 |
|
| 1112 |
/* |
| 1113 |
* When ending a parallel worker, copy the statistics gathered by the |
| 1114 |
* worker back into shared memory so that it can be picked up by the main |
| 1115 |
* process to report in EXPLAIN ANALYZE. |
| 1116 |
*/ |
| 1117 |
if (node->shared_info != NULL && IsParallelWorker()) |
| 1118 |
{ |
| 1119 |
MemoizeInstrumentation *si; |
| 1120 |
|
| 1121 |
/* Make mem_peak available for EXPLAIN */ |
| 1122 |
if (node->stats.mem_peak == 0) |
| 1123 |
node->stats.mem_peak = node->mem_used; |
| 1124 |
|
| 1125 |
Assert(ParallelWorkerNumber < node->shared_info->num_workers); |
| 1126 |
si = &node->shared_info->sinstrument[ParallelWorkerNumber]; |
| 1127 |
memcpy(si, &node->stats, sizeof(MemoizeInstrumentation)); |
| 1128 |
} |
| 1129 |
|
| 1130 |
/* Remove the cache context */ |
| 1131 |
MemoryContextDelete(node->tableContext); |
| 1132 |
|
| 1133 |
/* |
| 1134 |
* shut down the subplan |
| 1135 |
*/ |
| 1136 |
ExecEndNode(outerPlanState(node)); |
| 1137 |
} |
| 1138 |
|
| 1139 |
void |
| 1140 |
ExecReScanMemoize(MemoizeState *node) |
| 1141 |
{ |
| 1142 |
PlanState *outerPlan = outerPlanState(node); |
| 1143 |
|
| 1144 |
/* Mark that we must lookup the cache for a new set of parameters */ |
| 1145 |
node->mstatus = MEMO_CACHE_LOOKUP; |
| 1146 |
|
| 1147 |
/* nullify pointers used for the last scan */ |
| 1148 |
node->entry = NULL; |
| 1149 |
node->last_tuple = NULL; |
| 1150 |
|
| 1151 |
/* |
| 1152 |
* if chgParam of subnode is not null then plan will be re-scanned by |
| 1153 |
* first ExecProcNode. |
| 1154 |
*/ |
| 1155 |
if (outerPlan->chgParam == NULL) |
| 1156 |
ExecReScan(outerPlan); |
| 1157 |
|
| 1158 |
/* |
| 1159 |
* Purge the entire cache if a parameter changed that is not part of the |
| 1160 |
* cache key. |
| 1161 |
*/ |
| 1162 |
if (bms_nonempty_difference(outerPlan->chgParam, node->keyparamids)) |
| 1163 |
cache_purge_all(node); |
| 1164 |
} |
| 1165 |
|
| 1166 |
/* |
| 1167 |
* ExecEstimateCacheEntryOverheadBytes |
| 1168 |
* For use in the query planner to help it estimate the amount of memory |
| 1169 |
* required to store a single entry in the cache. |
| 1170 |
*/ |
| 1171 |
double |
| 1172 |
ExecEstimateCacheEntryOverheadBytes(double ntuples) |
| 1173 |
{ |
| 1174 |
return sizeof(MemoizeEntry) + sizeof(MemoizeKey) + sizeof(MemoizeTuple) * |
| 1175 |
ntuples; |
| 1176 |
} |
| 1177 |
|
| 1178 |
/* ---------------------------------------------------------------- |
| 1179 |
* Parallel Query Support |
| 1180 |
* ---------------------------------------------------------------- |
| 1181 |
*/ |
| 1182 |
|
| 1183 |
/* ---------------------------------------------------------------- |
| 1184 |
* ExecMemoizeEstimate |
| 1185 |
* |
| 1186 |
* Estimate space required to propagate memoize statistics. |
| 1187 |
* ---------------------------------------------------------------- |
| 1188 |
*/ |
| 1189 |
void |
| 1190 |
ExecMemoizeEstimate(MemoizeState *node, ParallelContext *pcxt) |
| 1191 |
{ |
| 1192 |
Size size; |
| 1193 |
|
| 1194 |
/* don't need this if not instrumenting or no workers */ |
| 1195 |
if (!node->ss.ps.instrument || pcxt->nworkers == 0) |
| 1196 |
return; |
| 1197 |
|
| 1198 |
size = mul_size(pcxt->nworkers, sizeof(MemoizeInstrumentation)); |
| 1199 |
size = add_size(size, offsetof(SharedMemoizeInfo, sinstrument)); |
| 1200 |
shm_toc_estimate_chunk(&pcxt->estimator, size); |
| 1201 |
shm_toc_estimate_keys(&pcxt->estimator, 1); |
| 1202 |
} |
| 1203 |
|
| 1204 |
/* ---------------------------------------------------------------- |
| 1205 |
* ExecMemoizeInitializeDSM |
| 1206 |
* |
| 1207 |
* Initialize DSM space for memoize statistics. |
| 1208 |
* ---------------------------------------------------------------- |
| 1209 |
*/ |
| 1210 |
void |
| 1211 |
ExecMemoizeInitializeDSM(MemoizeState *node, ParallelContext *pcxt) |
| 1212 |
{ |
| 1213 |
Size size; |
| 1214 |
|
| 1215 |
/* don't need this if not instrumenting or no workers */ |
| 1216 |
if (!node->ss.ps.instrument || pcxt->nworkers == 0) |
| 1217 |
return; |
| 1218 |
|
| 1219 |
size = offsetof(SharedMemoizeInfo, sinstrument) |
| 1220 |
+ pcxt->nworkers * sizeof(MemoizeInstrumentation); |
| 1221 |
node->shared_info = shm_toc_allocate(pcxt->toc, size); |
| 1222 |
/* ensure any unfilled slots will contain zeroes */ |
| 1223 |
memset(node->shared_info, 0, size); |
| 1224 |
node->shared_info->num_workers = pcxt->nworkers; |
| 1225 |
shm_toc_insert(pcxt->toc, node->ss.ps.plan->plan_node_id, |
| 1226 |
node->shared_info); |
| 1227 |
} |
| 1228 |
|
| 1229 |
/* ---------------------------------------------------------------- |
| 1230 |
* ExecMemoizeInitializeWorker |
| 1231 |
* |
| 1232 |
* Attach worker to DSM space for memoize statistics. |
| 1233 |
* ---------------------------------------------------------------- |
| 1234 |
*/ |
| 1235 |
void |
| 1236 |
ExecMemoizeInitializeWorker(MemoizeState *node, ParallelWorkerContext *pwcxt) |
| 1237 |
{ |
| 1238 |
node->shared_info = |
| 1239 |
shm_toc_lookup(pwcxt->toc, node->ss.ps.plan->plan_node_id, true); |
| 1240 |
} |
| 1241 |
|
| 1242 |
/* ---------------------------------------------------------------- |
| 1243 |
* ExecMemoizeRetrieveInstrumentation |
| 1244 |
* |
| 1245 |
* Transfer memoize statistics from DSM to private memory. |
| 1246 |
* ---------------------------------------------------------------- |
| 1247 |
*/ |
| 1248 |
void |
| 1249 |
ExecMemoizeRetrieveInstrumentation(MemoizeState *node) |
| 1250 |
{ |
| 1251 |
Size size; |
| 1252 |
SharedMemoizeInfo *si; |
| 1253 |
|
| 1254 |
if (node->shared_info == NULL) |
| 1255 |
return; |
| 1256 |
|
| 1257 |
size = offsetof(SharedMemoizeInfo, sinstrument) |
| 1258 |
+ node->shared_info->num_workers * sizeof(MemoizeInstrumentation); |
| 1259 |
si = palloc(size); |
| 1260 |
memcpy(si, node->shared_info, size); |
| 1261 |
node->shared_info = si; |
| 1262 |
} |